基于视觉图像的鲅鱼新鲜程度的检测方法

鲅鱼新鲜程度是评价其质量好坏的重要因素。为提高鲅鱼检测新鲜程度准确性，研究基于视觉图像的鲅鱼新鲜程度的检测方法。研究对象是某地海鲜市场中的30条鲅鱼，通过鲅鱼视觉图像采集系统采集鲅鱼视觉图像，利用区域填充算法及形态学开运算对采用大津法分割的鱼体二值图像进行填充及去噪，融合上山法与区域生长方法分割鱼眼区域，通过全局动态阈值分割方法分割鱼鳃图像；提取图像特征时，利用图像的R、G、I分量灰度均值提取鱼体、鱼眼及鱼鳃图像颜色特征，采用G分量提取鱼眼中心区域面积。将图像特征输入到NeuroShell 2神经网络判别模型中，实现鲅鱼新鲜程度的有效检测。经实验验证，该方法检测鲅鱼新鲜程度的准确率平均高达98.28%，依据鱼眼中心区域面积+颜色灰度均值特征进行鲅鱼新鲜程度检测的准确率最高，且检测不同死亡时间的鲅鱼新鲜度的检测准确率高达95%，说明鲅鱼新鲜度的检测为海鲜检测提供了理论基础。

Study on the Detection Method of Fresh Level of Spanish Mackerel Based on Visual Image

Freshness of Spanish mackerel is an important factor to evaluate its quality. In order to improve the accuracy of Spanish mackerel freshness detection, a visual image-based method was established. The research object of this work is 30 Spanish mackerel purchased from a seafood market. Spanish mackerel visual images were selected by an acquisition system. The region filling algorithm and morphological opening operation were used to fill and remove the binary image of fish segmented by Otsu method. Noise, merge the uphill method and region growing method were used to segment fish eye region. The global dynamic threshold segmentation method was used to segment fish gill image. When extracting image features, the color features of fish body, fish eye and fish gill image were extracted by using the gray mean of R, G and I components of image, and the area of fish eye center area was extracted by using G component. The image features were input into Neuro Shell 2 neural network discriminant model to detect freshness of marine Spanish mackerel effectively. The average accuracy of Spanish mackerel fresh method detection was 98.28%. The highest accuracy of detection of freshness was obtained by using the fisheye center area + color grayscale average characteristics of Spanish mackerel. And detecting the accuracy in determination of the freshness of Spanish mackerel of different death time, was s 95%. Results showed that this detection of freshness of Spanish mackerel could provide a theoretical basis for seafood detection.